CPU Temps in Linux

So, I’ve been running Manjaro Gnome on my Eurocom Sky X7C (i7 9700 KF, RTX 2080 @ 200 Watts, 16 GB DDR4, 1TB m.2) for almost a year now, and it seems to me like my CPU is running hot in Linux in general. Here’s what the core temperatures are like with just a single tab of Firefox running. Games would usually push things up to mid 70s to mid 80s. Today, after about 45 minutes of playing Rise of the Tomb Raider the CPU temps (on a specific core) was reaching up to 98 degrees!

Screenshot-from-2020-04-08-12-34-09

On Window$, which I would rather not have on my computer at all, the CPU temps seem to remain more or less normal, hovering around 45-55 degrees with similar usage. NZXT Cam is showing me 45 degrees on the CPU and 43 on the GPU with six tabs in FIrefox, Windows Mail, NZXT Cam and Steam running.

Does anyone have any idea as to what is causing this temperature difference? The CPU has not been overclocked or anything, even though I do have the unlocked BIOS, and I am wondering if this is some kind of bad microcode problem with Intel? Taking a look at system monitor doesn’t show anything with a major CPU workload either. The laptop is less than a year old, has upgraded copper heatsinks, delidded CPU and kryonaut on all surfaces… so I hardly think it’s a pasting issue.

This is a laptop, I wouldn’t expect desktop temps.

I’m not familiar with this specific laptop. I suppose the first thing to do would be to check if the fans are running properly.

When you’re running ROTTR, how hard was the CPU being hit?

I can’t really find great pictures of the cooling situation here, but I suspect it’s not perfect.

Now I don’t really know what’s going on with the temp differential between windows and Linux, but I suspect the voltage-frequency table for power states is a bit tighter than on Linux.

Did you do the delidding/kryonaut yourself? If it were me, I’d have a look, but that’s just me.


As far as ways to remedy this issue, I think undervolting is probably your best bet.

I use this on my XPS 15 to keep it under control:

Now, on my laptop I was able to get -150mv fully stable, but you may have different results. It will also allow you to limit the turbo situation, so you don’t turbo quite so high if that’s something you’re interested in.

Additionally, some Intel CPUs have temp sensors on the VRM. If the VRM is getting too toasty, it might throttle. I don’t know if it includes the VRM temps in “package temp” or not, but some do, some don’t.

If your VRM doesn’t have cooling on it, thermal pads to dump the heat into the case might be a good idea.


It’s also worth noting that we have system integrators here that have had horrible experiences with Clevo laptops. I will bring him into the thread to see if he’s got any advice.

For reference, the Clevo chassis is model P775TM1-G, according to my research.

Honestly, when it comes to laptops, 70s-80s is not abnormal in my experience. They tend to run warmer.

I think 98 indicates a thermal throttle, meaning cooling failure. Did the laptop vents have adequate airflow?

2 Likes

I am curious of the cooler. Is this a delidded desktop CPU or just a normal desktop CPU? What paste is on there?

I’m wondering if you have a bios update. Look and see and see if people have verified it as good. I unfortunately had to deal with a bad bios screwing with CPU temps and no GPU display, but luckily I can go back a bios version so no biggie. You might try out a different CPU paste. I don’t know if you’d want liquid metal, but I was surprised at how well it was working with a stock intel cooler recently… But you didn’t hear any recommendations from me.

Other thing I’d say to do is change your fans. If its- as thick of a laptop as it looks you can swap out what are probably 35CFM blowers for some 65CFM blowers out of an imac easy. I have done so, it came out amazing, but the fans also fit in the chassis. But don’t knock that as an option.

2 Likes

Fans are running fine, based on both the RPM counts and the acoustics. The delidding and pasting was all done by Eurocom, so my screwing things up is not a possibility.

I’ve usually heard good things about this particular Clevo chassis when it comes to cooling. Of course, I am not expecting desktop temperatures by any means… but I still think that 65-70 degrees for an idle machine is a little too hot. For comparison, I had left the laptop on for the past six hours on Window$, and both the CPU and GPU were idling comfortably at 38 degrees.

I will have to check the VRM thing… but again, I don’t see why that particular issue wouldn’t recur on Window$!

It’s a delidded desktop CPU… the delidding was done by Eurocom. All surfaces are using Thermal Grizzly Kryonaut.

I will have to look into the BIOS thing. Thanks for that! But once again I must wonder, if it was a BIOS issue it would probably hit Windows too, right? I expect there to be some difference between the two platforms, obviously, but I doubt that difference would be in the range of 30 degrees!

Interesting, Okay. Hmmm. Have you reached out to them to ask if they’ve had similar complaints?

You said it’s less than a year old, so while it’s obviously not ideal to need to send it back, you should get your complaints on record with them in case warranty work is needed after all.

Well, if you’re running higher voltages on linux because of different configuration (this would be default kernel configs, most likely), the VRM would be disproportionately affected.

I tend to agree. Even my XPS doesn’t idle that high.

Have you reached out to Eurocom?

So now that I think of it, Nvidia on Linux does use more energy than Windows. Something about the lowest power state being different. Are you using the GPU as always-on or is it a BumbleBee setup on Linux?

But, your GPU and CPU are on two entirely separate cooling systems, correct? (I don’t know for sure because the pictures I found never showed back off, heatsinks on)

Additionally, being a flagship of the current gen, the drivers are probably not mature. I like to stay 1 gen back on GPUs for Linux systems, but that’s just me I know you have high performance requirements.


EDIT: I just saw your location. It’s end of summer where you are, right? If so, could high ambient temps be a contributor?

Additionally, I’m making the assumption that you’re not resting this laptop on fabric, which would block the intakes.

1 Like

You said it’s less than a year old, so while it’s obviously not ideal to need to send it back, you should get your complaints on record with them in case warranty work is needed after all.

Good idea! I have now emailed them.

Always on… no optimus on this Clevo model. Plus the monitor is GSync enabled.

Yep… no shared heat pipes.

Funny thing though, the GPU temps on Linux, idle and under load, are comparable to Window$!!

Ah, okay.

Well, there goes that theory.

I suspect something wonky is going on with the CPU cooler. I’m going to reach out to the person I know who had a similar situation and get relevant information.

1 Like

It depends on the BIOS, but some set a preboot power limit using Intel RAPL and some don’t. The Windows drivers generally do it. Linux does not.

Which means that if the BIOS didn’t set the limit, the CPU is free to run boost all the way up to the 95C thermal limit.

I’d recommend finding a tool like rapl-set (or write your own little shell script) and tweaking the watt limit until you find one you’re happy with.

Oh yeah, this only affects high load temps, not idle. For idle temperatures you’re going to need to set the fan curve higher.

2 Likes

Okay, so.

Update: heard back from my contact. The issue he had was unrelated to cooling, however he mentioned that the Clevo systems need more thermal paste than the average system to make a solid connection.

1 Like

I am going to give this a try. Now that I think of it, the Clevo Control Center is used to set CPU/GPU/FAN/RGB modes in Window$ and sometimes on rebooting into linux it would fuck up the RGB settings. May be it’s screwing the fan settings too!

Thanks a lot Sgt! I will check the CPU paste and see if there’s something going on there. But again, a pasting issue should crop up in Windows too…

1 Like

I agree, which is why I suspect it’s got something to do with voltage at certain frequencies.

I’d give that github link I referenced above a try, it also has a feature where it lets you see current power usage in watts. It’s fairly detailed, so I think it might be able to help us pinpoint if the issue is package power or fan speed. Which, I suspect are the two potential culprits.

1 Like

So, when you boot into Linux, if the fan curve settings aren’t saved in the BIOS (most fan control systems don’t do this), it’ll fall back to the BIOS default curve, which is usually not what I’d consider acceptable. Now, the solution to this varies by motherboard (in your case, laptop), but if you install lm_sensors and fancontrol, you can set curves in Linux.

1 Like

Yeah… I’m trying that now. But for some reason fansensors-gui keeps telling me that there aren’t any pwm-capable fans in my system. I think I screwed something up while installing the packages may be…

No, probably not.

Sometimes you need to find and load a kernel module.

it’s nearly 1am here, so I can’t do much more tonight, but if you look around, you might be able to find your laptop documented somewhere.

Thanks! I will take a look! It’s around 3.44 am here…

How are you coping with the lockdown?

1 Like

I concur with using the Intel undervolt. Not sure exactly how low you can realistically go since it’s a desktop part.

Something else you could try is changing your governor setting. It could be that Linux is scaling the CPU more aggressively. There’s a few ways to do that, but I’ve used cpupower with some success.

https://wiki.archlinux.org/index.php/CPU_frequency_scaling

Between those two you should be able to get temps better than they are in windows stock.

2 Likes

To be honest, it’s not terribly difficult for me. I’ve been training for this my whole life. Ordering food by delivery, not leaving the house for weeks? The hardest part is not seeing friends, but I suspect this’ll be over in a couple months, and I can put up with nearly anything for 6 months.

1 Like

Okay… after first set of trials, neither fancontrol nor nbfc seems to work. With lmsensors and fancontrol I can’t get the software to detect any pwm-capable fans at all! Same with nbfc… in fact, a simple sensors-detect command is not returning any fan/rpm information at all. Just the core temperatures…